NVIDIA Releases New GPU Chip Tesla T4 with Turing Architecture

According to related reports, NVIDIA's Tesla T4 surpasses the RTX2070 and consumes only 75W. 

It handles data very fast, beyond our imagination. Huang referred to as the "greatest leap since the invention of the CUDA GPU in 2006."

The overview of NVIDIA Tesla T4

Recently, at the GTC 2018 graphics technology conference in Japan, NVIDIA released a new GPU Tesla T4, the first time using the same Turing Turing architecture as the RTX 20 series game card, the core specification surpasses the RTX 2070, the power consumption is incredibly low. 

Tesla T4 and TensorRT software are a new platform for AI applications, including intelligent voice, translation, video, image, recommendation and other applications. 

It claims to process queries 40 times faster than CPU only, with lower latency.

The card uses a PCI-E expansion card style, full coverage of the metal casing, the heat sink is hidden inside, the overall texture is quite capable, and there is no video output interface for the computing field.

It is equipped with 2,560 Turing architecture CUDA cores, 2944 lower than RTX 2080, but higher than RTX 2070 2304.

In addition, it can integrate 320 Tensor inference cores at the same time, but does not mention the RT ray tracing core, which may be blocked because it is useless.

In terms of performance, FP32 single-precision floating-point computing performance 8.1TFlops (8.1 trillion times per second).

FP16/32 half-precision and single-precision mixed computing performance 65TFlops, INT8, INT4 integer computing performance is 130Tops, 260Tops.

Tag: NVIDIA Releases New GPU Chip Tesla T4 with Turing Architecture

Previous: Qualcomm Announced The Launches Of First 5G Millimeter Wave Smartphone Antenna Module

Next: Ams a number of innovative sensor technologies appear in MWC